Abstract
2241
Objectives Age-matched normal FDG brain databases ideally created on the same PET scanner are required for diagnosis of different types of dementia using statistical mapping. However, the spatial distribution of the FDG uptake depends on the spatial normalization method. Aims of this study are a) to build normal FDG databases using a b-spline method (Lobster), SPM8, and Neurostat from a cohort of normal controls (NC) and b) to assess their suitability for clinical application.
Methods 37 NC (age: 51-85) were scanned on a Philips Gemini TF PET/CT scanner. Spatial normalization used the same FDG template and the creation of a normal FDG brain database was performed within each software package. For evaluation, 20 FDG brain studies of age-matched patients with Alzheimer’s disease (AD) were tested. For each method, scaling was performed to the median voxel value and z-score images (z= (NC-AD) / STDDEV(NC)) were calculated. Then, the cluster sizes at threshold: z>3 and maximum z-scores were extracted.
Results Comparing the NC databases, Lobster showed more FDG uptake in orbito-frontal cortex and in subependymal white matter than SPM and Neurostat (paired t-tests, p<0.0001). Testing the NC databases with FDG images of AD patients, in each AD scan Lobster, SPM and Neurostat identified typical patterns of hypometabolism in areas known to be affected in AD. However, Lobster showed larger clusters and higher z-scores (+10-15%) than SPM and Neurostat.
Conclusions Scanner-specific NC databases created with Lobster, SPM and Neurostat showed similar results in detecting reduced FDG uptake in AD patients. Higher z-scores and slightly larger clusters in Lobster depended upon methodological differences to SPM and Neurostat. The clinical relevance of differences between the NC databases created by Lobster, SPM and Neurostat has to be determined on the ground of larger cohorts